[Paddle-pipelines] Add pytorch retrieval model tutorials (#8159)

* Add pytorch retrieval model tutorials

* Update doc

* fix unitests

* Update docs

* Update torch2paddle tutorial

pipelines/FAQ.md

@@ -221,3 +221,48 @@ click版本过高导致：
 ```
 pip install click==8.0
 ```
+
+#### 怎么样新增最新的pytorch的检索模型
+
+PaddleNLP-Pipelines 提供了可自动将 PyTorch 相关的权重转化为 Paddle 权重的接口，以BAAI/bge-large-zh-v1.5为例，代码如下：
+
+```python
+from paddlenlp.transformers import AutoModel, AutoTokenizer
+model = AutoModel.from_pretrained("BAAI/bge-large-zh-v1.5", from_hf_hub=True, convert_from_torch=True)
+tokenizer = AutoTokenizer.from_pretrained('BAAI/bge-large-zh-v1.5', from_hf_hub=True)
+
+model.save_pretrained("BAAI/bge-large-zh-v1.5")
+tokenizer.save_pretrained("BAAI/bge-large-zh-v1.5")
+```
+
+然后在这里像这样注册一下即可使用：
+
+```
+"BAAI/bge-large-zh-v1.5": {
+                "task_class": SentenceFeatureExtractionTask,
+                "task_flag": "feature_extraction-BAAI/bge-large-zh-v1.5",
+                "task_priority_path": "BAAI/bge-large-zh-v1.5",
+            },
+```
+
+[taskflow 注册地址](https://github.com/PaddlePaddle/PaddleNLP/blob/b6dcb4e19efd85911b13a0fc587fef33578cfebf/paddlenlp/taskflow/taskflow.py#L680)
+
+使用方式示例如下：
+
+```
+document_store = FAISSDocumentStore.load(your_index_name)
+retriever = DensePassageRetriever(
+    document_store=document_store,
+    query_embedding_model="BAAI/bge-large-zh-v1.5",
+    passage_embedding_model="BAAI/bge-large-zh-v1.5",
+    output_emb_size=None,
+    max_seq_len_query=64,
+    max_seq_len_passage=256,
+    batch_size=16,
+    use_gpu=True,
+    embed_title=False,
+    pooling_mode="mean_tokens",
+)
+```
+
+**注意** bge-m3的底座模型是XLMRobertaModel，paddlenlp没有实现，不推荐使用。

pipelines/pipelines/document_stores/faiss.py

@@ -684,7 +684,7 @@ def _load_init_params_from_config(
                 init_params["index_name"] = []
                 for index in index_path:
                     faiss_index = faiss.read_index(str(index))
-                    index_name = str(index).split("/")[-1]
+                    index_name = os.path.basename(str(index))
                     init_params["index_name"].append(index_name)
                     init_params["faiss_index"][index_name] = faiss_index
                     # Add other init params to override the ones defined in the init params file

pipelines/tests/nodes/file_converter/test_pdf.py

@@ -25,10 +25,16 @@ def test_conversion(self):
 
         expected_result = [
             {
-                "content": " A Simple PDF File \n This is a small demonstration .pdf file - \n just for use in the Virtual Mechanics tutorials. More text. And more \n text. And more text. And more text. And more text. \n And more text. And more text. And more text. And more text. And more \n text. And more text. Boring, zzzzz. And more text. And more text. And \n more text. And more text. And more text. And more text. And more text. \n And more text. And more text. \n And more text. And more text. And more text. And more text. And more \n text. And more text. And more text. Even more. Continued on page 2 ...\x0c Simple PDF File 2 \n ...continued from page 1. Yet more text. And more text. And more text. \n And more text. And more text. And more text. And more text. And more \n text. Oh, how boring typing this stuff. But not as boring as watching \n paint dry. And more text. And more text. And more text. And more text. \n Boring.  More, a little more text. The end, and just as well. ",
+                "content": " A Simple PDF File \n This is a small demonstration .pdf file - \n just for use in the Virtual Mechanics tutorials. More text. And more \n text. And more text. And more text. And more text. \n And more text. And more text. And more text. And more text. And more \n text. And more text. Boring, zzzzz. And more text. And more text. And \n more text. And more text. And more text. And more text. And more text. \n And more text. And more text. \n And more text. And more text. And more text. And more text. And more \n text. And more text. And more text. Even more. Continued on page 2 ...",
                 "content_type": "text",
                 "meta": None,
-            }
+            },
+            {
+                "content": " Simple PDF File 2 \n ...continued from page 1. Yet more text. And more text. And more text. \n And more text. And more text. And more text. And more text. And more \n text. Oh, how boring typing this stuff. But not as boring as watching \n paint dry. And more text. And more text. And more text. And more text. \n Boring.  More, a little more text. The end, and just as well. ",
+                "content_type": "text",
+                "meta": None,
+            },
         ]
         result = converter.convert(file_path, process_num=1)
+        self.assertEqual(len(result), 2)
         self.assertEqual(expected_result, result)